Grammar of Graphics and Ggplot2

An influential book Grammar of Graphics by Leland Wilkinson lays out a theoretical framework to systematically construct quantitative graphics. Hadley Wickham extended this and related ideas and implemented them in the ggplot2 package. If you are interested in more details, this paper would be a great starting point.

At a very high level, this is what you would do with ggplot:

  • create a plot with the ggplot() call
  • add layers using various geom_*() or stat_*() calls. Some properties (e.g. position) of these layers can be set in these calls.
  • specify “aesthetics” using aes() for the full plot or per-layer. This deals with selecting variables to plot, colours, sizes, shapes of the plot elements etc. Aesthetics can be “mapped” to variables in the data.
  • optionally specify faceting
  • optionally adjust the scale, the axes and the legend
  • optionally customise the theme

Resources:

For a quick overview of ggplot, see https://r4ds.had.co.nz/data-visualisation.html.

To learn and practice more, the Rmd files here are great: https://github.com/hadley/ggplot2-book

Plotting the Urban Agglomerations

We load the data saved by the tidy.Rmd file (part-2 of the tutorial).

library(tidyverse)

data <- read_csv("data/ua_populations.csv")
sex_ratio_change <- read_csv("data/sex_ratio_change.csv")

Create a tidy population data set

pop_data <- data %>% 
  rename(total = population, male = pop_male, female = pop_female) %>% 
  select(-c(area, pop_change, pop_change_percent)) %>% 
  gather(sex, population, total, male, female)

pop_data
## # A tibble: 8,331 x 5
##    ua_no ua                   year sex   population
##    <dbl> <chr>               <dbl> <chr>      <dbl>
##  1     1 Greater Mumbai U.A.  1961 total    4515495
##  2     1 Greater Mumbai U.A.  1971 total    6596370
##  3     1 Greater Mumbai U.A.  1981 total    9421962
##  4     1 Greater Mumbai U.A.  1991 total   12596243
##  5     1 Greater Mumbai U.A.  2001 total   16434386
##  6     1 Greater Mumbai U.A.  2011 total   18394912
##  7     2 Delhi  U.A.          1961 total    2359408
##  8     2 Delhi  U.A.          1971 total    3647023
##  9     2 Delhi  U.A.          1981 total    5760811
## 10     2 Delhi  U.A.          1991 total    8471625
## # ... with 8,321 more rows

Male and Female Population Distributions

pop_data %>% 
  filter(sex != "total" & year == 1961) %>% 
  ggplot(aes(x = log(population), fill = sex, alpha = 0.5)) +
    geom_density(position = "identity")


Try changing geom_density() to geom_histogram(). Also try the position = "dodge" option. You could also add facet_wrap() to get the distributions for the years separately.

Total Population Distributions over Time

library(ggridges)

pop_data %>% 
  filter(sex == "total") %>% 
  ggplot(aes(x = log10(population), 
             y = factor(year, levels = rev(levels(factor(year)))))) +
    stat_density_ridges()


ggridges is a ggplot extension. There are many such useful extensions, and you can see the list here.

Line graphs - change over time

Plotting for one city - total population over time

pop_data %>%
  filter(ua == "Vapi ^" & sex == "total") %>% 
  ggplot(aes(x = year, y = population)) +
    geom_line()

Facet wrap

gap_increasing <- sex_ratio_change %>% 
  top_n(20, sex_ratio_slope)

gap_decreasing <- sex_ratio_change %>% 
  top_n(-20, sex_ratio_slope)

p <- pop_data %>%
  semi_join(gap_decreasing, by = "ua_no") %>% 
  filter(sex == "total") %>% 
  ggplot(aes(x = year, y = population)) +
    geom_line() +
    facet_wrap(~ua, scales = "free_y")

p + theme(axis.text.x = element_text(angle = 45, hjust = 1))

Multiple cities - male vs. female

Disparity closing

pop_data %>%
  semi_join(gap_decreasing, by = "ua_no") %>% 
  filter(sex != "total") %>% 
  ggplot() +
    geom_line(aes(x = year, y = population, colour = sex)) + 
    facet_wrap(~ua, scales = "free_y", ncol = 4, strip.position = "bottom") +
    theme(axis.text.x = element_text(angle = 45, hjust = 1))

Disparity increasing

pop_data %>%
  semi_join(gap_increasing, by = "ua_no") %>% 
  filter(sex != "total") %>% 
  ggplot() +
    geom_line(aes(x = year, y = population, colour = sex)) + 
    facet_wrap(~ua, scales = "free_y", ncol = 4, strip.position = "bottom") +
    theme(axis.text.x = element_text(angle = 45, hjust = 1))

2011 population: bar plot

pop_data %>% 
  filter(year == 2011 & sex == "total") %>% 
  arrange(desc(population)) %>% 
  top_n(30) %>% 
  ggplot() +
    geom_bar(aes(x = reorder(factor(ua), population), y = population), stat = "identity") +
    coord_flip() +
    xlab("City") +
    scale_y_continuous(label = scales::comma)

Scatter plot

cities_scatter <- data %>% 
  filter(year == 2011) %>% 
  mutate(
    sex_ratio = pop_male / pop_female,
    pop_density = population / area     
  ) %>%
  ggplot(aes(x = log10(population), y = pop_density, label = ua)) +
    geom_point(aes(size = log10(population),
                   colour = sex_ratio,
                   alpha = 0.4)) +
    scale_color_gradient(low = "green", high = "red") +
    scale_y_reverse()

cities_scatter

With labels

library(ggrepel)

cities_scatter_labelled <- cities_scatter +
  geom_label_repel(aes(label = ifelse(sex_ratio > quantile(sex_ratio, 0.98, na.rm = TRUE) |
                                        sex_ratio < quantile(sex_ratio, 0.02, na.rm = TRUE) |
                                        pop_density > quantile(pop_density, 0.98, na.rm = TRUE) |
                                        pop_density < quantile(pop_density, 0.02, na.rm = TRUE), 
                                      ua, '')))

cities_scatter_labelled

Interactive Scatter plot using plotly

Plotly is a library to create interactive visualisation. They have an API that follows the tidyverse philosophy. This book is a great resource to get started if you are interested. Plotly also provides a magic API to bring static ggplot plots to life. Here we will use only that.

plotly::ggplotly(cities_scatter)

Scatter Plot Animation

Let’s try something fancy!

Ever felt like doing a Hans Rosling? R lets you do some of that as well! In case you don’t know what I am talking about, whatch this video (seek to 11:00 if you’re in a hurry).

#library(animation)

mega_cities <- data %>% 
  filter(year == 2011 & population > 3000000 & ua_no != 6) %>% 
  select(ua_no)
  
draw_scatter <- function(y) {
  p <- data %>% 
    filter(year == y) %>% 
    semi_join(mega_cities, by = "ua_no") %>% 
    mutate(
      sex_ratio = pop_male / pop_female,
      pop_density = population / area     
    ) %>%
    ggplot(aes(x = sex_ratio, y = pop_density, label = ua)) +
      geom_point(aes(size = sqrt(population) / 300,
                     colour = ua,
                     alpha = 0.5)) +
      xlim(.8, 1.8) +
      ylim(25000, 0) +
      guides(size = FALSE, alpha = FALSE) + 
      scale_size_identity()
  
  ggsave(paste0("images/anim_", y, ".png"), plot = p)
  
  p
}

## saveGIF is not working for some reason. We'll save the png's and convert them to GIF using Imagemagick.
#saveGIF(c(1961, 1971, 1981, 1991, 2001, 2011) %>% map(get_scatter), 
#        interval = .2,
#        movie.name = "megacities_evolution.gif")

scatter_plots <- c(1961, 1971, 1981, 1991, 2001, 2011) %>% map(draw_scatter)

shell("magick convert -delay 100 -loop 1 images/anim*.png images/megacities_evolution.gif")

Here is the generated GIF file (to see the animation, refresh the page or open the image in a new window/tab on the browser)


Megacities

library(gganimate)

data %>% 
    semi_join(mega_cities, by = "ua_no") %>% 
    mutate(
      sex_ratio = pop_male / pop_female,
      pop_density = population / area     
    ) %>%
    ggplot(aes(x = sex_ratio, y = pop_density, label = ua)) +
      geom_point(aes(size = sqrt(population) / 300,
                     colour = ua,
                     alpha = 0.5)) +
      xlim(.8, 1.8) +
      ylim(25000, 0) +
      guides(size = FALSE, alpha = FALSE) + 
      scale_size_identity() +
      labs(title = 'Year: {frame_time}', x = 'Sex Ratio', y = 'Population Density') +
      transition_time(as.integer(year)) +
      ease_aes('linear')

R - Links and Resources